FS7: A million dumb basic questions about LUTs

[Liam Hall]

Sam, charts like that only have a limited use as there are so many elements that contribute to perceived sharpness, not just lens and aperture, but shutter, lighting, camera movement, colour rendition, codec and monitor to name a few obvious ones. I prefer to trust my eyes and experience.

 

[morgan_moore]

Sam, charts like that only have a limited use as there are so many elements that contribute to perceived sharpness, not just lens and aperture, but shutter, lighting, camera movement, colour rendition, codec and monitor to name a few obvious ones. I prefer to trust my eyes and experience.

Yep they have one purpose.. demonstration of micro detail. Here we can see the F5 falling apart at 3.4k, demonstrating the futility or worse of sending the camera a signal of more that 3.4k - if you dont like the rainbow effect one needs to keep the input signal below 3 k.

 

[starcentral]

To say that adjusting the black, mid, and white points of the red, green, and blue channels is the best form of basic color correction is to totally dismiss the value of color science, color management, and color processing of raw images. You are saying you can get better color with those old simple tools than by getting the gain and white balance right in your camera. You are saying that raw image processors need not bother with exposure or white balance sliders.

The argument is ridiculous on its face, and makes me think you've never really tried using color management or color science for color correction.

I have no doubt you are quite adept with the old simple tools, in large part because you have a lot of experience and you know how to get things almost perfect in the camera. I think you overestimate other people's abilities. I respect what you do and I think a lot of people here really value your opinions, but in this case I think this is not good advice for the average user.

For a guy with "a million dumb basic questions about LUTs", more advanced color correction is going to help him get better images with less effort. And after the basic color correction is done, you still have the option of adjusting the black, mid, and white points of the red, green, and blue channels, for example if you want to do different things to the shadows and highlights. But you shouldn't start with those adjustments. There's a reason every raw image processor has its exposure and white balance sliders way up at the top of the color correction controls. Logarist provides these basic controls with more ease and accuracy than the simple tools you're talking about. And it removes the complications of guessing which LUTs match your camera's color space, the difference between video levels and data levels, what to do with super-whites, and which color correction controls you should start with, because I've figured out all of that for you. And it's free, so why not try it? I made Logarist because I believe it's a better way of color correcting and a lot of people can benefit from it.

For you to even suggest color correction is not possible or that you can not white balance or set exposure by using color density based tools such as curves, lift/gamma/gain, 3-way tells me you honestly have no idea what you are talking about.. Nearly all, if not all professional coloring software tools are based on these "simple" underlying principles. This is strangely ironic coming from someone who developed a plugin for color correction and claims to know about color science and color management. So I guess you know more than Adobe, BlackMagic Design, Lightspace, etc.. ?

In a pinch if your tool works for people that's great but when asked I will always recommend that people take the time to learn the proper tools so that when they need to get further with their work they can. Same as myself and others will recommend people do not use LUTs for grading - just look at the issues it's causing Lenny who started this thread in the first place. Doing things the right way is called being smart and practical, it saves time in the long run. You already acknowledged that with your tool if its not enough someone can still go in later and adjust selective points across the various channels - if they are going to have to learn how to do that anyway then white balancing and exposure adjustment isn't realistically going to take a lot more knowledge. Please don't underestimate people's ability to learn, especially with community forums like this one where we all chime in and help one another. (not to mention all the free YouTube videos out there)


I tested your plugin and it horribly clipped my slog images from my Sony F55 compared to just putting the Sony LC709A LUT on it.

original slog3 image

Sony LC709A

logarist BT.709

I also tested your slider controls and found that you are sliding/skewing the entire channel data up and down right off the IRE scale. This will introduce channel clipping and as a result a loss of video information which is very bad for video. Especially for higher DR scenes or as elements within a scene change and light values move around. When you use proper tools based around lift/gamma/gain or curves data values are squashed or expanded as to preserve data which is the number one goal when performing color correction and especially when you are dealing with 8-bit or 10-bit video files in tiny Rec709 space that on top of it all is compressed with codec.

I can definitively say that your tool actually performs nothing like the exposure tools found in Adobe Camera RAW or Adobe Lightroom. In these professional tools data values are never truncated. I certainly hope more people become aware of the problems your plugin can create and are very careful using it, or as I had suggested learn how to use the proper tools.

And thanks for voting down all my posts trying to point out some simple and true facts.

 

[balazer]

The black point of S-Log3 is 3.5%. For your scopes to be showing values below 3.5% either something is wrong with the footage before the scope was applied, or it's an out-of-gamut very saturated dark green that's supposed to be gamut clipped in the gamut transformation. I can't say which without seeing actual images. Let me know if you want help understanding that.

I also tested your slider controls and found that you are sliding/skewing the entire channel data up and down right off the IRE scale. This will introduce channel clipping and as a result a loss of video information which is very bad for video.

It's not an IRE scale anymore after the Logarist input LUT is applied. It's a log color space, and it doesn't stop at 0.0 or 1.0. The Logarist color controls are appropriate for a log color space. It's very similar to ACEScc, if you've ever used that.

I can definitively say that your tool actually performs nothing like the exposure tools found in Adobe Camera RAW or Adobe Lightroom. In these professional tools data values are never truncated.

There are certainly cases where Adobe Camera Raw and Lightroom truncate. The output does not have unlimited dynamic range or gamut. The highlights get clipped at some point in the output rendering, and the gamut when you reduce to the size of the display gamut. That's how it's supposed to work. The idea is to record more picture information than you need, and decide later how to use all of it.

 

[moldcad]

Guys, I'm by far less knowledgeable in digital acquisition, color science etc. than (most of) you here, but I'd like to add just two observations to this interesting thread:

1. Sharpening.
Well - based on my experience with various 4K cams so far (am comparing the FS7, A7Rii and GH5 which I own) - sharpening in post is only required (and brings more good than harm to the picture) when:

a) the picture from the camera is soft to begin with (sounds obvious, eh?) Well - if any of the 3 cameras I mentioned ever gives a picture that calls for sharpening, it's only my FS7 (especially in Cine EI mode for obvious reasons). Both A7Rii and GH5 produce such a sharp picture with their default internal sharpening setting that during grade, adding more would look ridiculous (in fact, I tend to set internal sharpening to some minus values in these two cams)

b) when I zoom-into the 4K picture (either for cropping/panning/re-framing, or as a result of using strong stabilization settings). But even then, 0.46 is way too much even with the FS7 picture (ugly electronic edges instead of details); I usually leave it at 0.48 for the FS7 or 0.49 for the other cameras (plus the latter is very seldom needed)

2. Logarist workflow
As I mentioned several times elsewhere, i used to be so disappointed with many other LUT-based methods that for a long time my Resolve workflow was based 100% on RCM (with some ACES projects now and then, but mainly for experimentation purposes). After starting to use Logarist LUTs on input and output, I was amazed with 3 things:

a) the much nicer roll-off and color saturation of my highlights
b) the super-intuitive behavior of the color wheels in LOG mode
c) the easiness of matching the FS7 look not only with that of my A7Rii, but also the GH5

Just my $0.03; thank you

Piotr

 

[LennyLevy]

I just tried Logarist very briefly on the same clips I previously corrected in Lumetri with LUT's made by me in LUTCalc for Slog3 material. This is very preliminary but the basic LUT technique worked pretty well and I felt it got me into the right ball park faster than I did with the exposure modified LUT's I made in LUTCalc. I modified the initial Legalist corrections with various tools in Lumetri (especially Saturation) and very quickly had a look similar to what I did previously. Certainly no clipping but the clips I used didn't have a full range of 14 stops.

 

[morgan_moore]

Piotr

All my comments about sharpening have referred to shooting EI which disables sharpening in camera.

I am suspicious that even setting sharpening to 0 on "baked" cameras properly disables it.

My 46-47 comment was from memory 47-49 May be better

 

[moldcad]

Piotr

All my comments about sharpening have referred to shooting EI which disables sharpening in camera.

I am suspicious that even setting sharpening to 0 on "baked" cameras properly disables it.

My 46-47 comment was from memory 47-49 May be better

Sure you're right, Sam - the FS7 softness is the most striking in CineEI where you can't crank the sharpness up in the camera (I have edited my previous post to reflect this). But even in Custom, with HG7 I use sometimes in low-light situations, increasing the internal sharpness in the Picture Profile will never give me amount of fine detail comparable with either GH5 or A7Rii (in their default settings).

Piotr

 

[morgan_moore]

Hg settings are reasonably grown up. The consumer cams are dialled to jump off the screen in a manner that we find as putrid as a cheap consumer monitor in "image enhancement" mode

 

[moldcad]

True about the "image enhancement" settings of prosumer devices (my Samsung UHD TV even has "shop display mode" in the menu, haha) - but the sharpness of the A7Rii is simply higher than FS7's; you can see it even with detail dialed down in the former's and up in the latter's PP...

Piotr

 

[stokstad]

3.4k is pretty ideal IMO I just wish we could get a proper 2K 12bit 4:4:4 like on the c300mk2

 

[Nick_Shaw]

The black point of S-Log3 is 3.5%. For your scopes to be showing values below 3.5% either something is wrong with the footage before the scope was applied, or it's an out-of-gamut very saturated dark green that's supposed to be gamut clipped in the gamut transformation.

In fact the black point of S-Log3 is only 3.5% if you interpret the footage as "Video". If you interpret it as "Data", which is what should happen with an S-Log3 MXF automatically, the black point is 95/1023 which is about 9.3%. The ACES IDT as well as the standard Sony LC709A LUT expect input which has been interpreted as "Data". External recordings such as ProRes from an Odyssey have no metadata to tell Resolve whether to interpret as "Video" for baked Rec.709 recordings or "Data" for S-Log ones. "Auto" will just set the ProRes default of "Video", resulting in a black level which is at 3.5%. That is why the LC709A LUTs from Convergent Design have been modified (by me) to work correctly with image data interpreted this way.

 

[balazer]

Nick, in this case we are talking about Premiere, not Resolve.

 

[Nick_Shaw]

Nick, in this case we are talking about Premiere, not Resolve.

Ah, the joy of the same clip being interpreted differently in different apps!

 

[LennyLevy]

Can anyone suggest a good place to read a simple primer(s) that explains Data and Legal levels with regard to use of LUT's , also something related differences between Premiere and Resolve, and also re use of ACES

 

[morgan_moore]

Is legal an outmoded concept? There is various debate on this.

Len - first you see here people have opinions - all somewhat funny - which is why I suggested stick with a Sony issued lut.

If you want to know about srgb and rec709 wiki is a good place to start.

Aces is well laid out on the academy website - what is not clarified is tha you need a 10bit wide gamut momitor to do it properly!

 

[Nick_Shaw]

Is legal an outmoded concept? There is various debate on this.

The complications of full/legal range (and some of the issues are caused by software trying to hide the complications from users) may be frustrating, but it's the situation we are in, and we need to deal with it. Signals travel down SDI cables in legal range, because they are usually Y'CbCr, which is by definition legal range. ProRes and other formats are internally legal range. So "outmoded" or not it exists, and we need to deal with it.

Len - first you see here people have opinions - all somewhat funny - which is why I suggested stick with a Sony issued lut.

The Sony LUT is only appropriate if the image data is correctly scaled for the input to that LUT. It would appear that Premiere's range interpretation means it isn't. So in this case, the Sony LUT is inappropriate. The output of Sony's LC709A LUT will not match the LC709A from the camera, because the input to the LUT is wrong.

Aces is well laid out on the academy website - what is not clarified is tha you need a 10bit wide gamut momitor to do it properly!

You need a monitor which is calibrated to your target deliverable. It does not need to be wide gamut unless you are delivering in a wide gamut format such as Rec.2020. 10-bit is always ideal (although unrelated to whether you are using ACES or not) but plenty of things have been graded on 8-bit monitors.

If you are delivering "Rec.709" you want a good reference monitor calibrated to BT.1886. ACES changes nothing in that regard.

 

[morgan_moore]

Not arguing. Chatting/Learning.

SDI.. Is the OPs file going down SDI in his workflow at all , if it is actually broadcast would not the broadcaster legalise before transmission? For web only.. then no SDI

The Sony LUT is only appropriate if the image data is correctly scaled for the input to that LUT.

Indeed, I suggest LC709 - which is this for?

(ACES) You need a monitor which is calibrated to your target deliverable.

My (wooly understanding) is that you grade on a high spec monitor and the data/gamut/gamma is compressed to the deliverable 'width' by the ODT (output display transform) upon export

The whole purpose of ACES (according to my wooly understanding) is to grade once for many deliverables not to grade to the gamut/gamma etc of the deliverable.. giving a huge time saving and look consistency to a production that needs to produce different deliverables (phone, apple trailers, broadcast, cinema, future unknown device)

So you exactly dont grade on a monitor of your target deliverable, although a test would of course be prescient to check the ODT is functioning properly.


S

 

[broughtonfilm]

Indeed, I suggest LC709 - which is this for?

I think you missed the point, he's talking about the image data of the clip, no LUT is used yet. It does not matter what you use and how appropriate the LUT is or not if that is not correct first.

 

[LennyLevy]

Nick, It's very nice to see someone with your background in grading and LUT's here - welcome and thanks for helping out.

I'm doing better than I was when I originally posted here and have been getting good looking grades but feel like I'm making it up as I go and don't have a solid basis. I'm way behind on understanding what you guys are talking about here but I'm wondering if anyone can help with an understanding of the choices LUTCalc offers.

You can make a LUT that is Data In Data Out, Data In Legal Out, or L in L out and L in D out.
You also have choices about whether you want your LUT to have hard clips at the bottom or the top but I don't know why you would do that in a LUT instead of just doing it in Post later.

I found for Premiere that Data in Legal Out gave me most dynamic range in the LUT itself which is really the only reason i chose to work with those. (Maybe these question should go to Ben though. Ben does partially explain this issue in LUTCalc instructions but not enough for my understanding which to use. He suggested experimenting for Premiere which is what I did but is the dynamic range of the LUT all i should care about?)

Also it seems to be taken for granted that Premiere and Resolve handle LUT's differently. Can anyone explain that difference?

Still focusing on Premiere BTW because my clients are using it to cut and grade and I need to understand how to work in their worlds first.